The 4 Paid Social Lessons AI Still Can't Learn

AI can automate reporting, recommendations, and optimizations. It can't replace the judgment required to diagnose pacing issues, identify creative failures, challenge platform automation, and tell a business story with data.

While at Publicis, I helped oversee more than 100 live paid social campaigns at once. Was this our recommendation? No. But did we execute it? Yes.

I worked with Taco Bell, Jared Jewelers, and Universal Destinations & Experiences. Six platforms. National flights layered on regional promos layered on always-on. A brand team, creative agency, influencer team, platform reps, third-party measurement partners, and a client who wanted one number on one slide.

Here are four things that stuck from those years, and why they matter more now with ~800 AI tools for this job.

#1: Pacing is a knowledge gap before it’s a math problem.

Any tool can tell you you’re 12% under-delivered. What it can’t tell you is that it’s because your creative refresh is stuck with legal, the Q4 promo calendar shifted, or the landing page is broken. You need someone who’s seen it before — someone who can tell you if it’s a normal platform fluctuation, something worth digging into, or a real problem and exactly where to start.

#2: creative that doesn’t look native will always underperform

I’ve seen teams pour budget into creative that looks like an obvious ad on every platform. It doesn’t blend, it doesn’t stop the scroll — and then someone blames the products featured on the landing page.

But the products weren’t the problem. The target market wasn’t pictured in the creative. Without someone who’s diagnosed this before, the team optimizes the wrong thing.

This is where AI helps — and where it needs an expert. AI can analyze your audience, study native content on each platform, and recommend concepts that fit. But it needs someone steering the brief, because AI will happily generate beautiful creative that completely misses who your customer is.

#3: platforms’ “smart” tools aren’t always the smart choice

Every platform is leaning into AI: smart audiences, smart placements, smart budgets. But “smart” doesn’t mean right for your campaign.

Select a smart audience with a narrow target and specific message, and you’ve handed budget to an algorithm optimizing for volume, not relevance. Choose smart placements and your creative gets served where it doesn’t fit the specs — looking wonky in placements nobody clicks on.

Knowing when to trust the automation and when to override it is judgment you only build by being wrong a few times on each platform.

#4: Reporting is translation, not export

The hardest deliverable every month wasn’t the data pull. It was turning six platforms, 100+ campaigns and three objectives into something a VP could repeat in a meeting. AI is phenomenal at the pull and the first draft; it still needs someone who knows which number the room cares about.

Have questions? Let’s talk.

Next
Next

AI Tools for Paid Social: Put Workflow Expertise to the Test